Founder & Principal Software Architect

Scoria Software Solutions | March 2025 – Present

Role Summary

Founder-level architecture work focused on deterministic AI systems, model-agnostic orchestration, Azure Databricks-backed sports analytics, and production-grade tooling for software, document, data, infrastructure, and DevOps workflows. The record is strongest where a reviewer needs evidence of AI safety guardrails, measurable workflow compression, token and cloud-cost reduction, modern analytics architecture, and multi-platform delivery discipline.

• Designed every tool to operate inside explicit, deterministic guardrails with codified approval and immutable attribution for any divergence, producing clearer audit trails, faster incident response (seconds or milliseconds instead of minutes), and a 1–2 order-of-magnitude reduction in attempts to bypass safety controls.
• Built service-principal lifecycle tooling with ephemeral-token preference, check-in/check-out workflows, granular least privilege, and agentic session-ID attribution, reducing non-human identity risk while preserving auditable customer delivery.
• Reduced false positives by more than 50% (reaching over 95% in mature deployments) and cut human-intervention escalations from 20–30 per month to roughly one per quarter at one client, freeing security specialists to resume long-shelved SASE work.
• Built the XLM engine and MCP toolkit family that turns a single prompt into a complete, deterministic, multi-platform application stack while remaining fully model- and platform-agnostic, enabling early adopters to collapse POC cycles from months to weeks, double win rates, and increase inbound leads tenfold.
• Made the suite model- and platform-agnostic so the same tools run interchangeably across consumer apps, enterprise platforms, Azure AI Foundry, Ollama, and local models; one customer cut Azure spend 65% while dropping application latency from 3–5 seconds to 10–150 ms.
• Engineered identity-provider-agnostic access tooling that can operate across Entra-native development, CyberArk, HashiCorp, and other privileged-identity planes, keeping customer controls portable instead of binding governance to one vendor interface.
• Architected Fumaro Sports with Azure Databricks as the analytics backend showcase, using lakehouse-oriented data engineering patterns to support predictive sports intelligence, model-ready feature preparation, context-aware analytics, and governed human-in-the-loop release controls.
• Designed Fumaro Sports authentication for B2C customer access, SSO, and enterprise tenancy so organizations can run governed, tenant-aware versions of the model and agent experience without collapsing consumer and enterprise identity boundaries.
• Designing Fumaro Sports around analyst and scout review, feedback capture, telemetry-aware runtime logging, staged private beta evaluation, and human-in-the-loop release controls so sports tools can improve without hiding uncertainty from users.
• Created the DevOps orchestration engine with built-in AI-development guardrails that reduced technical debt 65–95% of the codebase and compressed feature and patch delivery from weeks or months to days or hours, supporting daily stable releases at one client.
• Implemented Microsoft Graph and Azure Management API-native control-plane automation for Entra and Microsoft Cloud operations, enabling agentic workflows with stronger audit visibility and finer least-privilege boundaries than default portal-driven administration.
• Developed a single canonical JSON schema front-end engine that deploys cohesive applications across web, native mobile, desktop, APIs, CLIs, and MCP toolkits, reducing each interface’s codebase 25–30% through shared components and enabling one team to manage all interfaces instead of dedicated teams per platform.
• Built the vendor- and OS-agnostic infrastructure engine spanning Azure, AWS, private cloud, bare metal, and OpenStack, enabling consistent portable deployments and stronger vendor negotiation leverage that contributed to the observed 65% cloud-cost reduction.
• Engineered the back-end data engine for live, byte-perfect interconversion and normalization across major relational, document, and file formats, allowing agents to work exclusively in JSON and eliminating the token overhead normally spent on formatting and translation.
• Designed and implemented a proprietary Rust conversion engine delivering byte-perfect bidirectional Markdown/MDX ↔ DOCX/PDF translation, eliminating branding and styling errors and collapsing white-paper and sales-collateral production from 1–2 weeks to 1–2 days (more than half same-day including human review).
• Reduced token consumption in LLM-driven document generation by 60–90% across four workflow steps, enabling models to produce publication-ready output with zero post-processing.
• Marketed and deployed the suite to technology companies, MSPs, and consulting firms, enabling functional hands-on POCs before competitors could pitch, single-source vendor status on multiple public-sector bids, 20–40% lower delivery cost to end customers without eroding margins, and significantly accelerated revenue recognition.
• Led scalable cloud-solution and software-delivery architecture for client environments, connecting security assessment, DevOps, data, and application patterns so customers could improve operational efficiency without separating modernization from risk control.
• Conducted cybersecurity architecture and assessment work for client systems handling sensitive data, translating compliance and protection requirements into deployable patterns rather than abstract policy language.
• Coordinated cross-functional design and delivery across software, cloud, data, infrastructure, and security workstreams, giving clients a single accountable architecture thread from concept through production adoption.